[lsb-discuss] Multi-lib standards

[Tobin Davis]

From: Wichmann, Mats D <mats.d.wichmann at intel.com>
To: Anthony W. Youngman <wol at thewolery.demon.co.uk>,
lsb-discuss at lists.linux-foundation.org
<lsb-discuss at lists.linux-foundation.org>
Subject: Re: [lsb-discuss] Multi-lib standards
Date: Wed, 2 Dec 2009 10:06:40 -0700

Anthony W. Youngman wrote:
> In message <4B0AE280.8050801 at gmail.com>, Bruce Dubbs
> <bruce.dubbs at gmail.com> writes
>> tytso at mit.edu wrote:
>> 
>>> And given that there are some Atom chips which are 32-bit only, and
>>> a number of people who tend to run 32-bit VM's (because they want to
>>> conserve memory), there will be many ISV's who may very well decide
>>> they only want to ship 32-bit products, just to simplify their
>>> build/test matrices.
>> 
>> Of course there should be continued support for 32-bit systems.  The
>> vast majority of computer users don't *need* 64-bit capability, but
>> many think they do. 
>> 
>> I wonder how much memory users actually save by running in 32-bit
>> mode on 64-bit systems.  Memeory is pretty cheap, but there are also
>> many who persist on running on 512M when memory is under $40/G.
> 
> PLEASE DON'T assume that just because you can afford a humungous
> system, that others can too.
> 
> Oh - and your argument that memory is cheap is TOTALLY SPECIOUS. I'm
> typing this on a "more than powerful enough" system that has
> (admittedly more than 512Mb) 768Mb. But to upgrade my RAM I will need
> a new motherboard to put it in, which means a new CPU, and a new hard
> drive, and probably a new power supply and/or case - in short,
> basically, I need a new computer! (To give you an idea, the bios
> copyright is 1999 and the processor is a Socket A Athlon).

>As if if was needed, I'll add one more point here:  we're moving
>more and more into an era where some people will have "consumer" devices
>which effectively aren't modifiable from their factory state at all -
>consider the (non-exclusive) list of netbooks/nettops/mids which
>either have no expansion capacity at all, or have expansion which
>requires removing a "this will void your warranty" sticker.

To add to this, Arm systems are on the rise.  Google's Android Phone OS
is getting a lot of good reviews and is rapidly improving (I own a
Verizon Motorola Droid).  I would venture to say that there soon will be
more Arm based systems than S390 systems very soon (how many distro's
support S390?).  At the Ubuntu Developers Summit in Dallas, we even saw
the first functional multi-core Arm server.  The company publicly
demoing it was running with 4 cores, and consuming somewhere around 35W.
Two other companies were publicly demoing arm based netbooks, with 8+
hours of battery life (one of which is already selling in Japan).

I am tasked with seeing about adding arm support to some of the LSB test
suite, just for API and ABI compatibility as we enable the new armv7
instruction set, which all of these systems are based on.  Any pointers
would be greatly appreciated.